There is academic, commercial, and public interest in estimating loss from hurricanes striking land and understanding how loss might change as a result of future variations in climate. We show in a paper in to be published in Geophysical Research Letters that the relationship between wind speed and loss is exponential and that loss increases with wind speed at a rate of 5% per m/s. The relationship is derived using quantile regression and a data set comprising wind speeds of hurricanes hitting the United States and normalized economic losses. We suggest that the “centercepts” for the different quantiles account for exposure-related factors such as population density, precipitation, and surface roughness, and that once these effects are accounted for, the increase in loss with wind speed is consistent across quantiles. An out-of-sample test of this relationship correctly predicts economic losses from Hurricane Irene in 2011. The exponential relationship suggests that increased wind speeds will produce significantly higher losses; however, increases in exposed property and population are expected to be a more important factor for near future losses. The research was directed by Richard Murnane of the Risk Prediction Initiative. The code used to obtain the results is available at http://rpubs.com/jelsner/816.
Friday, August 03, 2012
Maximum wind speeds and U.S. hurricane losses
There is academic, commercial, and public interest in estimating loss from hurricanes striking land and understanding how loss might change as a result of future variations in climate. We show in a paper in to be published in Geophysical Research Letters that the relationship between wind speed and loss is exponential and that loss increases with wind speed at a rate of 5% per m/s. The relationship is derived using quantile regression and a data set comprising wind speeds of hurricanes hitting the United States and normalized economic losses. We suggest that the “centercepts” for the different quantiles account for exposure-related factors such as population density, precipitation, and surface roughness, and that once these effects are accounted for, the increase in loss with wind speed is consistent across quantiles. An out-of-sample test of this relationship correctly predicts economic losses from Hurricane Irene in 2011. The exponential relationship suggests that increased wind speeds will produce significantly higher losses; however, increases in exposed property and population are expected to be a more important factor for near future losses. The research was directed by Richard Murnane of the Risk Prediction Initiative. The code used to obtain the results is available at http://rpubs.com/jelsner/816.